【摘要】 利用类间的散布矩阵,寻找特征脸子空间,让每一类在子空间中散布得更开;然后结合BP神经网络学习能力强、分类能力强的优点,利用它实现分类器。具体是将所有的样本投影到特征脸子空间中,并将每一个样本得到的特征系数作为BP神经网络的输入。实验证明,这种方法是有效的。更多还原

【Abstract】 The scatter matrix between classes is used to find the subspace. So the classes scatter more widely each other in the subspace. BP-based neural network is used as classifier, because it has a good learning capability.Firstly all samples are projected into the subspace, and then the feature coefficients of every sample are inputted to BP neural network for training. Experimental results demonstrate the method is effective.更多还原